This study is designed to establish the efficacy of the ketogenic diet in a blinded, placebo-controlled fashion. Twenty to thirty percent of children with epilepsy remain refractory to currently available anticoveulsant medications. The frequent atonic-myoclonic seizures of the Lennox-Gastaut Syndrome (LGS), are particularly handicapping. Preliminary date suggest that even when other medications have failed, these seizures respond rapidly and dramatically to a high-fat-low-carbohydrate ketogenic diet. Twenty children with LGS having more 15 atonic/ myoclonic seizures/day will enter a randomized, blinded, crossover protocol to evaluate the efficacy of this diet. Efficacy will be defined as greater than 50% decrease from baseline seizure frequency utilizing continuous EEG monitoring obtained just prior to fasting. The diet will be instituted in a standard fashion following 36 hours of fasting. Children will be randomized to a liquid supplement containing either placebo (maintaining ketosis) or glucose negating ketosis). Seizure frequency (utilizing 24 hours continuous EEG monitoring) will be evaluated after full diet initiation (5 days). The child will be re-fasted for 24 hours and crossed to the alternate liquid supplement. At the end of 5 days on the second arm, a repeat 24 hours continuous EEG monitoring will be obtained, and the child will be discharged on the routine ketogenic diet. The primary endpoint will be the number of EEG documented seizures on the last day of each treatment period. We do not believe that there will be substantive sequence or carry- over effect. However, these assumptions will be examined, first by comparing the seizure frequency after the placebo diet to the frequency measured at baseline, and then by calculating the total seizure frequency for each sequence and comparing the two sequences. After discharge, each child's diet will be "fine-tuned" to optimize seizure control, followed in routine fashion with monthly seizure calenders maintained by the parents. Each child will return for follow-up at six months and one year. Twenty-four hour continuous EEG monitoring will be obtained at each of these visits. The secondary question of whether control can be maintained overtime will consist of either a simple proportion (with confidence interval), or will use survival methods, if those are indicated because of poor compliance or loss to follow-up. Analysis will be done both by intention to treat and by those on treatment.